Electrical Conductivity of Mica in Intense Fields. 117 



had previously survived, an occurrence which is possibly o£ 

 some interest in connexion with theories of polarization. 

 To prevent this from happening again the resistances were 

 ultimately connected as shown in fig. 1, thus preventing 

 surges either at charge or discharge. 



The discharge-key, shown diagrammatically at D in fig. 1, 

 was formed of stout copper wires dipping into mercury cups 

 in a block of paraffin-wax. The wires were mechanically 

 connected by a rod of sealing-u ax. A rubber spring held 

 the key in the discharge position, but the key could be held 

 in the charge position, as shown in the figure, by means of a 

 trigger, on release of which the key very rapidly flew over 

 to the discharge position. As the voltages employed ex- 

 ceeded 5000, it was found necessary to make the gap at the 

 key at least a couple of centimetres long, as otherwise the 

 mercury sometimes splashed up far enough to enable a spark 

 to pass, thus discharging the large leyden-jar into one of the 

 galvanometers. As it was necessary that as short a time as 

 possible should elapse between the disconnexion of the con- 

 densers from the source and their connexion to the galvano- 

 meters, the mercury cups connected to the source were made 

 about 5 cm. deep and filled with mercury to within about 

 2 cm. of their tops, which were only about 0'6 cm. wide. 

 The wires of the key, when in the charge position, reached 

 nearly to the bottom of the cups, so that, on releasing the 

 trigger, the key had moved about 3 cm._, and thus had 

 acquired a considerable velocity, before it broke connexion 

 with the source. This made it possible to combine rapidity 

 of action with a sufficiently wide gap. The mouths of the 

 mercury cups were made narrow to reduce splashing. 



As the leakage through the mica is relatively large at the 

 higher pressures, it is necessary to ascertain whether a per- 

 ceptible loss of charge would occur during the time of 

 operation of the key. To estimate this interval a standard 

 •J microfarad condenser was charged to a suitable pressure 

 and discharged by means of the key through a ballistic gal- 

 vanometer. A resistance of 0*106 megohm was then con- 

 nected as a shunt across the condenser and the reading- 

 repeated. The leak of the condenser through the given 

 resistance in the interval between disconnexion from the 

 source and discharge, is measured b}^ the ratio of the two 

 throws. Measurements made in this way gave values of 

 this interval ranging from 00070 to 0*0093 sec. for the side 

 of the key connected to the mica condenser. The figure 

 0*008 was taken as a mean, and a very small correction pro- 

 portional to the leakage current was added to the observed 



